Classification automations for Slack
Turn team criteria into Slack actions.
Choose the channels. Describe what a message means. Let Jev select your outcomes and trigger the reactions, replies, and notifications you configured.
14-day free trial. From $10 per workspace per month. Model processing included.
Jev for SlackHomeInteractive example - simulated results, no Slack or model connection
First-line support triage
Message
Prompt
{{message_text}}
Outcomes & actions
React with 👀 · Reply: “Thanks, the support team will help here.”
Post message link to #support-escalations
Message
Prompt
{{message_text}}
Outcomes & actions
Post message link to #billing-help
Post message link to #checkout-feedback
Post message link to #identity-help
Message
Prompt
{{message_text}}
Outcomes & actions
React with ⚡
Post message link to #product-ideas
Reply with the access-help template
First-line support triage · Yes
Reaction added. Configured reply posted in the original thread.
Feedback signals · Performance, Feature request
Probabilities and rendered actions previewed. No Slack actions executed.
Product area routing
A transient provider failure delayed classification. The original 14-day deadline still applies.
Specialist notification · Yes
The channel post could not be confirmed. Review Slack before taking further action; this post was not blindly retried.
Four illustrative entries, not customer data. In the app, activity is visible for 14 days.
Manage workspace processing
A workspace administrator or owner authorizes external processing before rules or tests can run. Prompts, outcome descriptions and their referenced Slack values are sent to OpenRouter and TypeSafe.
This illustration does not change consent. In the app, withdrawing authorization stops new model requests and is checked before each action.
Read the data processing guideInteractive illustration with synthetic messages and scores. Select a classification mode or App Home section to explore. No Slack actions or model requests.
Your context
One message. A clear question.
Use explicit Slack variables to choose what reaches the model. There is no hidden whole-workspace context.
Your decisions
Outcomes you can explain.
Define yes/no, single-choice or multiple-choice answers. Give each one the criteria your team works by.
Your workflow
Actions you can preview.
Jev picks from your configured outcomes. Your templates determine the exact message or reaction that follows.
From conversation to action
Give your team’s criteria a place to run.
Is this a routine support request? Which product area should see this feedback? Does a message contain more than one useful signal? Turn recurring decisions into rules, right where the conversation happens.
Choose the conversation
Select channels accessible to both Jev and the rule owner and listen for new messages, edits, or both.
Describe the distinction
Write a question and explain what each answer means. Insert message context from the variable picker.
Close the loop
Attach actions to answers, preview a real message, and review results in a 14-day activity log.
How it works
Three steps to a rule.
A preview before you rely on it.
Explore the native Slack modal workflow. These are illustrative examples, with synthetic messages and results.
Start at 25%. Name the automation, select channels and message triggers, choose future-only behavior, and write a question using explicit Slack variables.
Create automation
Apply only to messages created after this rule
{{message_text}}At 50%, choose Yes / No, single choice, or multiple choice. Describe every answer so Jev knows which criteria to apply. Multiple choice supports a probability threshold.
Create automation
Routine how-to questions, password resets and known issues covered by our support documentation.
Engineering changes, new features or incidents that need specialist investigation.
Each answer has its own description. For multiple choice, each outcome is selected independently against your threshold.
At 75%, attach reactions, thread replies, or channel messages to the answers you defined. Back keeps your draft. Save automation enables the rule.
Create automation
Routine questions covered by our support documentation.
Needs specialist investigation.
{{message_link}}Paste a Slack message URL to test a saved rule. Review model probabilities, chosen answers, and rendered actions. A dry test does not execute any Slack actions.
Test automation
Planned actions for Yes
- 👀Add reaction
eyes · original message in #help-desk
- ↳Reply in thread
Thanks for the question. Our support team will help you in this thread.
Illustrative interface, not an embedded Slack session. Use the tabs above to explore all four views.
Explicit Slack context
Use the message values your rule needs.
Insert variables into your prompt and outcome descriptions with the help picker. Values used only in action text stay out of the classification request. Unknown variables fail validation.
{{message_text}}{{message_link}}{{channel_name}}{{channel_id}}{{user_id}}{{message_ts}}{{thread_ts}}{{workspace_name}}Three ways to act on an outcome
Action text comes from your templates. Jev can respond to a conversation without changing what someone else wrote.
Add a reaction
Add a configured emoji reaction to the source message, such as eyes or white_check_mark.
Reply in a thread
Post your configured response in the message thread. Use Slack variables to include relevant context.
Post to a channel
Send a configured message to an accessible destination channel, with a link back to the original conversation.
The app does not edit other people’s messages. Slack access and channel membership still apply.
Processing under your control
Your workspace decides when processing starts.
Explicit authorization
A workspace administrator or owner authorizes external model processing in Data processing. Consent is checked before classification and each action. Withdrawing it cannot recall a request already sent.
Selected channels
Rules run in channels selected by an administrator and accessible to the bot. Direct messages and bot messages are excluded. Administration is checked on the server.
Configured model input
The prompt, outcome descriptions and referenced Slack values go to OpenRouter and TypeSafe. Input exceeding the complete 30,000-byte request budget fails with guidance, without silent truncation.
A fixed retention window
Activity, pending work and action receipts expire 14 days after the original receipt. Retries do not extend the deadline. Rules have separate retention and remain until deleted or uninstalled.
The app runs on Cloudflare Workers with D1 storage. AI processing takes place outside Slack. Review security controls and data processing details.
Activity and recovery
See the decision. Follow the result.
Activity log brings the message reference, rule, scores, selected outcomes and action results together. Model probabilities are estimates, not a guarantee that a decision is correct.
14 days of activity
Review automatic runs and dry tests without storing full message bodies, expanded action text, or credentials in activity.
Recovery after an outage
Captured, eligible events wait for service recovery and retry transient failures with backoff, within their original deadline.
Retry with current checks
Retry now requests an immediate attempt. Recovery rechecks consent, rule revision, enabled state and current message visibility.
Disabling, editing or deleting a rule invalidates its queued work. Recovery cannot reconstruct events never delivered to the app.
Built for workspace administrators.
native App Home
Three focused sections
Manage Classification automations, inspect Activity log, and control workspace Data processing inside Slack.
yes/no · single · multi
Your choice of answers
Choose a yes/no decision, one named outcome, or multiple independent outcomes with a probability threshold.
created · edited · both
Channels and triggers
Choose the channels and message events for each rule. Future-only rules use the original message timestamp.
message URL → preview
Test before taking action
Paste an accessible Slack message URL. See probabilities, selected outcomes, and rendered actions without posting or reacting.
explicit variables
Only the context you choose
Reference Slack values in prompts and outcome descriptions. Action-only variables are expanded inside the app.
14-day recovery
Pick up after an outage
Eligible pending classifications retry after transient failures. Review the queue or request Retry now in Activity log.
Simple workspace pricing
One workspace.
Your team’s automations.
$5USD per active human member
per month
No workspace minimum.
Add to Slack14-day free trial. Payment method required. Activate from Slack App Home.
Jev processing is included.
- Publisher-funded model usage. No API key to supply.
- Yes/no, single-choice and multiple-choice automations.
- Reactions, thread replies and channel messages.
- Dry tests, a 14-day activity log and recovery controls.
- Workspace-isolated rules, activity and processing consent.
Active human members include guests. Bots and deactivated accounts are excluded. Each workspace gets one trial and its own Stripe subscription. Member counts refresh daily; changes affect the next invoice without proration. Cancel in the billing portal before uninstalling. Review billing terms at checkout.
Frequently asked
The questions admins ask first.
How do I get Jev for Slack?
Use Add to Slack to authorize Jev for your workspace, then activate the service in App Home. Your workspace may require administrator approval. Jev for Slack is independently published; no Slack Marketplace approval is claimed.
How much does it cost?
The price is USD $5 per active human workspace member per month, with no workspace minimum and a 14-day free trial. Active human guests count; bots and deactivated accounts do not. Each workspace gets one trial on completion of Stripe Checkout with a payment method. Cancel billing through the portal before uninstalling; uninstall does not cancel the subscription. We cover model usage, so you do not need an AI account or API key.
What can a rule do?
Each selected outcome can add a reaction, reply in the source thread, or post a configured message to a channel. Message text comes from your templates. Jev does not edit another person’s message or invent actions.
Who can manage automations?
Workspace administrators and owners manage rules, dry tests, activity, and processing consent from Jev’s App Home. The app checks administrative access on the server.
Does Jev read every conversation?
No. Rules select specific channels accessible to the bot. Direct messages and bot messages are not observed. Model requests use the prompt, outcome descriptions, and explicitly referenced Slack variables; there is no implicit whole-workspace context.
What happens when I save a rule?
Saving enables it for the selected future message events. It does not backfill old messages. The future-only setting compares the original message timestamp with the rule’s creation time, so editing an older message does not make it new.
What does a test do?
A test sends the configured input from an accessible message to the model and shows probabilities, chosen outcomes, and fully rendered actions. It does not post messages or add reactions. Workspace processing consent, active billing or trial, and access to a message in a channel selected by the rule are required.
What happens during an outage?
Captured, eligible events can wait for recovery for up to 14 days from their original receipt. Transient failures retry with backoff. Retry now requests another attempt; it does not bypass permissions, processing consent, or provider availability.
Will edited messages create duplicate replies?
Within their 14-day retention window, persisted event and action receipts prevent repeated events and message edits from intentionally stacking the same completed action. This protection is not permanent. If a post may have succeeded but cannot be confirmed, it requires review instead of a blind retry.
What leaves Slack?
The prompt, outcome criteria and their referenced Slack values are sent through OpenRouter to TypeSafe’s Jev model. The app is hosted on Cloudflare Workers with D1 storage. External providers have their own processing and retention terms.
What is retained?
Activity, pending recovery and action receipts have a fixed 14-day deadline from the original event receipt. Retries do not extend it. Activity and recovery logs exclude full message bodies and expanded action text. Temporary dry-test modal results can contain rendered action text, including message content, for up to 30 minutes. Rules remain until deleted or uninstalled; billing history is separate.
Can I use my own outcome criteria?
Yes. Write a description for every outcome, including both Yes and No. The three-step wizard separates the question, answer definitions, and actions. A variable help picker explains each supported Slack token.
Jev for Slack
A clearer next step for every conversation.
Install Jev in your workspace, activate your 14-day trial in App Home, and build your first classification rule. Your workspace keeps its own rules, activity and processing consent.
USD $5 per active human member per month, with no workspace minimum after the trial. Jev for Slack is independently published. Read the App Terms and Privacy Policy.